Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!wuarchive!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!ucbvax!agate!darkstar!ucsc.edu
From: taff@ucsc.edu (Tom Affinito)
Newsgroups: comp.lang.functional
Subject: TIM and SECD
Keywords: TIM, SECD
Message-ID: <16372@darkstar.ucsc.edu>
Date: 30 May 91 23:50:48 GMT
References: <949@macuni.mqcc.mq.oz>
Sender: usenet@darkstar.ucsc.edu
Lines: 24

Question from a novice:

I'm trying to get an intuition for Fairburn & Wray's TIM machine. I've read  
their '87 paper, and am familar with SECD, SKI reduction, and the G-machine.

What I'm trying to get is some deep feeling on how TIM relates to SECD and SKI  
reduction. It seems to me that TIM is very similar to what compiled code for an  
SECD system would be like....a series of instructions for manipulating the  
stacks, building environments, and storing partial results. Is the difference  
solely in the fact that environments need not be created in TIM because of the  
preliminary lambda-lifting phase? IE, if SECD had had a supercombinator  
preprocessor, might it have looked like TIM?

In what ways do SECD + graph reduction = TIM? In what ways should that = be <>?
(Abstractly of course)

Any flames, impressions, or weird musings are most welcome!!	:)

Tom Affinito
taff@kate.ucsc.edu
-----------------------------------------------
function for an arbitrary permutation of a vector argument in J:
p =. @:~?&!&#
Wow!